1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
5
3
# This program is free software; you can redistribute it and/or modify
6
4
# it under the terms of the GNU General Public License as published by
7
5
# the Free Software Foundation; either version 2 of the License, or
8
6
# (at your option) any later version.
10
8
# This program is distributed in the hope that it will be useful,
11
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
11
# GNU General Public License for more details.
15
13
# You should have received a copy of the GNU General Public License
16
14
# along with this program; if not, write to the Free Software
17
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
import os, types, re, time, errno, sys
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from errors import bailout, BzrError
23
from trace import mutter
17
from cStringIO import StringIO
21
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
22
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
43
from tempfile import (
55
# sha is deprecated in python2.6 but hashlib is available as of 2.5
56
if sys.version_info < (2, 5):
69
from bzrlib import symbol_versioning
70
from bzrlib.symbol_versioning import (
74
from bzrlib.trace import mutter
77
# On win32, O_BINARY is used to indicate the file should
78
# be opened in binary mode, rather than text mode.
79
# On other platforms, O_BINARY doesn't exist, because
80
# they always open in binary mode, so it is okay to
81
# OR with 0 on those platforms
82
O_BINARY = getattr(os, 'O_BINARY', 0)
26
85
def make_readonly(filename):
27
86
"""Make a filename read-only."""
28
# TODO: probably needs to be fixed for windows
29
mod = os.stat(filename).st_mode
31
os.chmod(filename, mod)
87
mod = os.lstat(filename).st_mode
88
if not stat.S_ISLNK(mod):
90
os.chmod(filename, mod)
34
93
def make_writable(filename):
35
mod = os.stat(filename).st_mode
37
os.chmod(filename, mod)
40
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/_~-])')
94
mod = os.lstat(filename).st_mode
95
if not stat.S_ISLNK(mod):
97
os.chmod(filename, mod)
100
def minimum_path_selection(paths):
101
"""Return the smallset subset of paths which are outside paths.
103
:param paths: A container (and hence not None) of paths.
104
:return: A set of paths sufficient to include everything in paths via
105
is_inside_any, drawn from the paths parameter.
110
other_paths = paths.difference([path])
111
if not is_inside_any(other_paths, path):
112
# this is a top level path, we must check it.
113
search_paths.add(path)
42
"""Return shell-quoted filename"""
43
## We could be a bit more terse by using double-quotes etc
44
f = _QUOTE_RE.sub(r'\\\1', f)
51
mode = os.lstat(f)[ST_MODE]
121
"""Return a quoted filename filename
123
This previously used backslash quoting, but that works poorly on
125
# TODO: I'm not really sure this is the best format either.x
127
if _QUOTE_RE is None:
128
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
130
if _QUOTE_RE.search(f):
59
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
136
_directory_kind = 'directory'
139
stat.S_IFDIR:_directory_kind,
140
stat.S_IFCHR:'chardev',
141
stat.S_IFBLK:'block',
144
stat.S_IFLNK:'symlink',
145
stat.S_IFSOCK:'socket',
149
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
150
"""Generate a file kind from a stat mode. This is used in walkdirs.
152
Its performance is critical: Do not mutate without careful benchmarking.
155
return _formats[stat_mode & 0170000]
160
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
162
return _mapper(_lstat(f).st_mode)
164
if getattr(e, 'errno', None) == errno.ENOENT:
165
raise errors.NoSuchFile(f)
170
"""Return the current umask"""
171
# Assume that people aren't messing with the umask while running
172
# XXX: This is not thread safe, but there is no way to get the
173
# umask without setting it
181
_directory_kind: "/",
183
'tree-reference': '+',
62
187
def kind_marker(kind):
65
elif kind == 'directory':
67
elif kind == 'symlink':
70
raise BzrError('invalid file kind %r' % kind)
189
return _kind_marker_map[kind]
191
raise errors.BzrError('invalid file kind %r' % kind)
194
lexists = getattr(os.path, 'lexists', None)
198
stat = getattr(os, 'lstat', os.stat)
202
if e.errno == errno.ENOENT:
205
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
208
def fancy_rename(old, new, rename_func, unlink_func):
209
"""A fancy rename, when you don't have atomic rename.
211
:param old: The old path, to rename from
212
:param new: The new path, to rename to
213
:param rename_func: The potentially non-atomic rename function
214
:param unlink_func: A way to delete the target file if the full rename succeeds
217
# sftp rename doesn't allow overwriting, so play tricks:
219
base = os.path.basename(new)
220
dirname = os.path.dirname(new)
221
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
222
tmp_name = pathjoin(dirname, tmp_name)
224
# Rename the file out of the way, but keep track if it didn't exist
225
# We don't want to grab just any exception
226
# something like EACCES should prevent us from continuing
227
# The downside is that the rename_func has to throw an exception
228
# with an errno = ENOENT, or NoSuchFile
231
rename_func(new, tmp_name)
232
except (errors.NoSuchFile,), e:
235
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
236
# function raises an IOError with errno is None when a rename fails.
237
# This then gets caught here.
238
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
241
if (getattr(e, 'errno', None) is None
242
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
249
# This may throw an exception, in which case success will
251
rename_func(old, new)
255
# If the file used to exist, rename it back into place
256
# otherwise just delete it from the tmp location
258
unlink_func(tmp_name)
260
rename_func(tmp_name, new)
263
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
264
# choke on a Unicode string containing a relative path if
265
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
267
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
268
def _posix_abspath(path):
269
# jam 20060426 rather than encoding to fsencoding
270
# copy posixpath.abspath, but use os.getcwdu instead
271
if not posixpath.isabs(path):
272
path = posixpath.join(getcwd(), path)
273
return posixpath.normpath(path)
276
def _posix_realpath(path):
277
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
280
def _win32_fixdrive(path):
281
"""Force drive letters to be consistent.
283
win32 is inconsistent whether it returns lower or upper case
284
and even if it was consistent the user might type the other
285
so we force it to uppercase
286
running python.exe under cmd.exe return capital C:\\
287
running win32 python inside a cygwin shell returns lowercase c:\\
289
drive, path = _nt_splitdrive(path)
290
return drive.upper() + path
293
def _win32_abspath(path):
294
# Real _nt_abspath doesn't have a problem with a unicode cwd
295
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
298
def _win98_abspath(path):
299
"""Return the absolute version of a path.
300
Windows 98 safe implementation (python reimplementation
301
of Win32 API function GetFullPathNameW)
306
# \\HOST\path => //HOST/path
307
# //HOST/path => //HOST/path
308
# path => C:/cwd/path
311
# check for absolute path
312
drive = _nt_splitdrive(path)[0]
313
if drive == '' and path[:2] not in('//','\\\\'):
315
# we cannot simply os.path.join cwd and path
316
# because os.path.join('C:','/path') produce '/path'
317
# and this is incorrect
318
if path[:1] in ('/','\\'):
319
cwd = _nt_splitdrive(cwd)[0]
321
path = cwd + '\\' + path
322
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
324
if win32utils.winver == 'Windows 98':
325
_win32_abspath = _win98_abspath
328
def _win32_realpath(path):
329
# Real _nt_realpath doesn't have a problem with a unicode cwd
330
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
333
def _win32_pathjoin(*args):
334
return _nt_join(*args).replace('\\', '/')
337
def _win32_normpath(path):
338
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
342
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
345
def _win32_mkdtemp(*args, **kwargs):
346
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
349
def _win32_rename(old, new):
350
"""We expect to be able to atomically replace 'new' with old.
352
On win32, if new exists, it must be moved out of the way first,
356
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
358
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
359
# If we try to rename a non-existant file onto cwd, we get
360
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
361
# if the old path doesn't exist, sometimes we get EACCES
362
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
368
return unicodedata.normalize('NFKC', os.getcwdu())
371
# Default is to just use the python builtins, but these can be rebound on
372
# particular platforms.
373
abspath = _posix_abspath
374
realpath = _posix_realpath
375
pathjoin = os.path.join
376
normpath = os.path.normpath
379
dirname = os.path.dirname
380
basename = os.path.basename
381
split = os.path.split
382
splitext = os.path.splitext
383
# These were already imported into local scope
384
# mkdtemp = tempfile.mkdtemp
385
# rmtree = shutil.rmtree
387
MIN_ABS_PATHLENGTH = 1
390
if sys.platform == 'win32':
391
abspath = _win32_abspath
392
realpath = _win32_realpath
393
pathjoin = _win32_pathjoin
394
normpath = _win32_normpath
395
getcwd = _win32_getcwd
396
mkdtemp = _win32_mkdtemp
397
rename = _win32_rename
399
MIN_ABS_PATHLENGTH = 3
401
def _win32_delete_readonly(function, path, excinfo):
402
"""Error handler for shutil.rmtree function [for win32]
403
Helps to remove files and dirs marked as read-only.
405
exception = excinfo[1]
406
if function in (os.remove, os.rmdir) \
407
and isinstance(exception, OSError) \
408
and exception.errno == errno.EACCES:
414
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
415
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
416
return shutil.rmtree(path, ignore_errors, onerror)
417
elif sys.platform == 'darwin':
421
def get_terminal_encoding():
422
"""Find the best encoding for printing to the screen.
424
This attempts to check both sys.stdout and sys.stdin to see
425
what encoding they are in, and if that fails it falls back to
426
bzrlib.user_encoding.
427
The problem is that on Windows, locale.getpreferredencoding()
428
is not the same encoding as that used by the console:
429
http://mail.python.org/pipermail/python-list/2003-May/162357.html
431
On my standard US Windows XP, the preferred encoding is
432
cp1252, but the console is cp437
434
output_encoding = getattr(sys.stdout, 'encoding', None)
435
if not output_encoding:
436
input_encoding = getattr(sys.stdin, 'encoding', None)
437
if not input_encoding:
438
output_encoding = bzrlib.user_encoding
439
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
441
output_encoding = input_encoding
442
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
444
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
445
if output_encoding == 'cp0':
446
# invalid encoding (cp0 means 'no codepage' on Windows)
447
output_encoding = bzrlib.user_encoding
448
mutter('cp0 is invalid encoding.'
449
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
452
codecs.lookup(output_encoding)
454
sys.stderr.write('bzr: warning:'
455
' unknown terminal encoding %s.\n'
456
' Using encoding %s instead.\n'
457
% (output_encoding, bzrlib.user_encoding)
459
output_encoding = bzrlib.user_encoding
461
return output_encoding
464
def normalizepath(f):
465
if getattr(os.path, 'realpath', None) is not None:
469
[p,e] = os.path.split(f)
470
if e == "" or e == "." or e == "..":
473
return pathjoin(F(p), e)
476
@deprecated_function(zero_ninetythree)
478
"""Copy a file to a backup.
480
Backups are named in GNU-style, with a ~ suffix.
482
If the file is already a backup, it's not copied.
488
if has_symlinks() and os.path.islink(fn):
489
target = os.readlink(fn)
490
os.symlink(target, bfn)
498
outf = file(bfn, 'wb')
271
647
def local_time_offset(t=None):
272
648
"""Return offset of local zone from GMT, either at present or at time t."""
273
# python2.3 localtime() can't take None
277
if time.localtime(t).tm_isdst and time.daylight:
280
return -time.timezone
283
def format_date(t, offset=0, timezone='original'):
284
## TODO: Perhaps a global option to use either universal or local time?
285
## Or perhaps just let people set $TZ?
286
assert isinstance(t, float)
651
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
652
return offset.days * 86400 + offset.seconds
655
def format_date(t, offset=0, timezone='original', date_fmt=None,
657
"""Return a formatted date string.
659
:param t: Seconds since the epoch.
660
:param offset: Timezone offset in seconds east of utc.
661
:param timezone: How to display the time: 'utc', 'original' for the
662
timezone specified by offset, or 'local' for the process's current
664
:param show_offset: Whether to append the timezone.
665
:param date_fmt: strftime format.
288
667
if timezone == 'utc':
289
668
tt = time.gmtime(t)
291
670
elif timezone == 'original':
294
673
tt = time.gmtime(t + offset)
295
674
elif timezone == 'local':
296
675
tt = time.localtime(t)
297
676
offset = local_time_offset(t)
299
bailout("unsupported timezone format %r",
300
['options are "utc", "original", "local"'])
302
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
303
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
678
raise errors.BzrError("unsupported timezone format %r" % timezone,
679
['options are "utc", "original", "local"'])
681
date_fmt = "%a %Y-%m-%d %H:%M:%S"
683
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
686
return (time.strftime(date_fmt, tt) + offset_str)
306
689
def compact_date(when):
307
690
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
693
def format_delta(delta):
694
"""Get a nice looking string for a time delta.
696
:param delta: The time difference in seconds, can be positive or negative.
697
positive indicates time in the past, negative indicates time in the
698
future. (usually time.time() - stored_time)
699
:return: String formatted to show approximate resolution
705
direction = 'in the future'
709
if seconds < 90: # print seconds up to 90 seconds
711
return '%d second %s' % (seconds, direction,)
713
return '%d seconds %s' % (seconds, direction)
715
minutes = int(seconds / 60)
716
seconds -= 60 * minutes
721
if minutes < 90: # print minutes, seconds up to 90 minutes
723
return '%d minute, %d second%s %s' % (
724
minutes, seconds, plural_seconds, direction)
726
return '%d minutes, %d second%s %s' % (
727
minutes, seconds, plural_seconds, direction)
729
hours = int(minutes / 60)
730
minutes -= 60 * hours
737
return '%d hour, %d minute%s %s' % (hours, minutes,
738
plural_minutes, direction)
739
return '%d hours, %d minute%s %s' % (hours, minutes,
740
plural_minutes, direction)
312
743
"""Return size of given open file."""
313
744
return os.fstat(f.fileno())[ST_SIZE]
316
if hasattr(os, 'urandom'): # python 2.4 and later
747
# Define rand_bytes based on platform.
749
# Python 2.4 and later have os.urandom,
750
# but it doesn't work on some arches
317
752
rand_bytes = os.urandom
318
elif sys.platform == 'linux2':
319
rand_bytes = file('/dev/urandom', 'rb').read
321
# not well seeded, but better than nothing
326
s += chr(random.randint(0, 255))
753
except (NotImplementedError, AttributeError):
754
# If python doesn't have os.urandom, or it doesn't work,
755
# then try to first pull random data from /dev/urandom
757
rand_bytes = file('/dev/urandom', 'rb').read
758
# Otherwise, use this hack as a last resort
759
except (IOError, OSError):
760
# not well seeded, but better than nothing
765
s += chr(random.randint(0, 255))
770
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
772
"""Return a random string of num alphanumeric characters
774
The result only contains lowercase chars because it may be used on
775
case-insensitive filesystems.
778
for raw_byte in rand_bytes(num):
779
s += ALNUM[ord(raw_byte) % 36]
331
783
## TODO: We could later have path objects that remember their list
332
784
## decomposition (might be too tricksy though.)
334
786
def splitpath(p):
335
"""Turn string into list of parts.
341
>>> splitpath('a/./b')
343
>>> splitpath('a/.b')
345
>>> splitpath('a/../b')
346
Traceback (most recent call last):
348
BzrError: ("sorry, '..' not allowed in path", [])
350
assert isinstance(p, types.StringTypes)
787
"""Turn string into list of parts."""
788
assert isinstance(p, basestring)
352
790
# split on either delimiter because people might use either on
367
assert isinstance(p, list)
805
assert isinstance(p, (list, tuple))
369
if (f == '..') or (f == None) or (f == ''):
370
bailout("sorry, %r not allowed in path" % f)
371
return os.path.join(*p)
374
def appendpath(p1, p2):
378
return os.path.join(p1, p2)
381
def extern_command(cmd, ignore_errors = False):
382
mutter('external command: %s' % `cmd`)
384
if not ignore_errors:
385
bailout('command failed')
807
if (f == '..') or (f is None) or (f == ''):
808
raise errors.BzrError("sorry, %r not allowed in path" % f)
813
"""Split s into lines, but without removing the newline characters."""
814
lines = s.split('\n')
815
result = [line + '\n' for line in lines[:-1]]
817
result.append(lines[-1])
821
def hardlinks_good():
822
return sys.platform not in ('win32', 'cygwin', 'darwin')
825
def link_or_copy(src, dest):
826
"""Hardlink a file, or copy it if it can't be hardlinked."""
827
if not hardlinks_good():
828
shutil.copyfile(src, dest)
832
except (OSError, IOError), e:
833
if e.errno != errno.EXDEV:
835
shutil.copyfile(src, dest)
838
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
839
# Forgiveness than Permission (EAFP) because:
840
# - root can damage a solaris file system by using unlink,
841
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
842
# EACCES, OSX: EPERM) when invoked on a directory.
843
def delete_any(path):
844
"""Delete a file or directory."""
845
if isdir(path): # Takes care of symlinks
852
if getattr(os, 'symlink', None) is not None:
858
def contains_whitespace(s):
859
"""True if there are any whitespace characters in s."""
860
# string.whitespace can include '\xa0' in certain locales, because it is
861
# considered "non-breaking-space" as part of ISO-8859-1. But it
862
# 1) Isn't a breaking whitespace
863
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
865
# 3) '\xa0' isn't unicode safe since it is >128.
867
# This should *not* be a unicode set of characters in case the source
868
# string is not a Unicode string. We can auto-up-cast the characters since
869
# they are ascii, but we don't want to auto-up-cast the string in case it
871
for ch in ' \t\n\r\v\f':
878
def contains_linebreaks(s):
879
"""True if there is any vertical whitespace in s."""
887
def relpath(base, path):
888
"""Return path relative to base, or raise exception.
890
The path may be either an absolute path or a path relative to the
891
current working directory.
893
os.path.commonprefix (python2.4) has a bad bug that it works just
894
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
898
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
899
' exceed the platform minimum length (which is %d)' %
906
while len(head) >= len(base):
909
head, tail = os.path.split(head)
913
raise errors.PathNotChild(rp, base)
921
def safe_unicode(unicode_or_utf8_string):
922
"""Coerce unicode_or_utf8_string into unicode.
924
If it is unicode, it is returned.
925
Otherwise it is decoded from utf-8. If a decoding error
926
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
927
as a BzrBadParameter exception.
929
if isinstance(unicode_or_utf8_string, unicode):
930
return unicode_or_utf8_string
932
return unicode_or_utf8_string.decode('utf8')
933
except UnicodeDecodeError:
934
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
937
def safe_utf8(unicode_or_utf8_string):
938
"""Coerce unicode_or_utf8_string to a utf8 string.
940
If it is a str, it is returned.
941
If it is Unicode, it is encoded into a utf-8 string.
943
if isinstance(unicode_or_utf8_string, str):
944
# TODO: jam 20070209 This is overkill, and probably has an impact on
945
# performance if we are dealing with lots of apis that want a
948
# Make sure it is a valid utf-8 string
949
unicode_or_utf8_string.decode('utf-8')
950
except UnicodeDecodeError:
951
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
952
return unicode_or_utf8_string
953
return unicode_or_utf8_string.encode('utf-8')
956
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
957
' Revision id generators should be creating utf8'
961
def safe_revision_id(unicode_or_utf8_string, warn=True):
962
"""Revision ids should now be utf8, but at one point they were unicode.
964
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
966
:param warn: Functions that are sanitizing user data can set warn=False
967
:return: None or a utf8 revision id.
969
if (unicode_or_utf8_string is None
970
or unicode_or_utf8_string.__class__ == str):
971
return unicode_or_utf8_string
973
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
975
return cache_utf8.encode(unicode_or_utf8_string)
978
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
979
' generators should be creating utf8 file ids.')
982
def safe_file_id(unicode_or_utf8_string, warn=True):
983
"""File ids should now be utf8, but at one point they were unicode.
985
This is the same as safe_utf8, except it uses the cached encode functions
986
to save a little bit of performance.
988
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
990
:param warn: Functions that are sanitizing user data can set warn=False
991
:return: None or a utf8 file id.
993
if (unicode_or_utf8_string is None
994
or unicode_or_utf8_string.__class__ == str):
995
return unicode_or_utf8_string
997
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
999
return cache_utf8.encode(unicode_or_utf8_string)
1002
_platform_normalizes_filenames = False
1003
if sys.platform == 'darwin':
1004
_platform_normalizes_filenames = True
1007
def normalizes_filenames():
1008
"""Return True if this platform normalizes unicode filenames.
1010
Mac OSX does, Windows/Linux do not.
1012
return _platform_normalizes_filenames
1015
def _accessible_normalized_filename(path):
1016
"""Get the unicode normalized path, and if you can access the file.
1018
On platforms where the system normalizes filenames (Mac OSX),
1019
you can access a file by any path which will normalize correctly.
1020
On platforms where the system does not normalize filenames
1021
(Windows, Linux), you have to access a file by its exact path.
1023
Internally, bzr only supports NFC/NFKC normalization, since that is
1024
the standard for XML documents.
1026
So return the normalized path, and a flag indicating if the file
1027
can be accessed by that path.
1030
return unicodedata.normalize('NFKC', unicode(path)), True
1033
def _inaccessible_normalized_filename(path):
1034
__doc__ = _accessible_normalized_filename.__doc__
1036
normalized = unicodedata.normalize('NFKC', unicode(path))
1037
return normalized, normalized == path
1040
if _platform_normalizes_filenames:
1041
normalized_filename = _accessible_normalized_filename
1043
normalized_filename = _inaccessible_normalized_filename
1046
def terminal_width():
1047
"""Return estimated terminal width."""
1048
if sys.platform == 'win32':
1049
return win32utils.get_console_size()[0]
1052
import struct, fcntl, termios
1053
s = struct.pack('HHHH', 0, 0, 0, 0)
1054
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1055
width = struct.unpack('HHHH', x)[1]
1060
width = int(os.environ['COLUMNS'])
1069
def supports_executable():
1070
return sys.platform != "win32"
1073
def supports_posix_readonly():
1074
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1076
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1077
directory controls creation/deletion, etc.
1079
And under win32, readonly means that the directory itself cannot be
1080
deleted. The contents of a readonly directory can be changed, unlike POSIX
1081
where files in readonly directories cannot be added, deleted or renamed.
1083
return sys.platform != "win32"
1086
def set_or_unset_env(env_variable, value):
1087
"""Modify the environment, setting or removing the env_variable.
1089
:param env_variable: The environment variable in question
1090
:param value: The value to set the environment to. If None, then
1091
the variable will be removed.
1092
:return: The original value of the environment variable.
1094
orig_val = os.environ.get(env_variable)
1096
if orig_val is not None:
1097
del os.environ[env_variable]
1099
if isinstance(value, unicode):
1100
value = value.encode(bzrlib.user_encoding)
1101
os.environ[env_variable] = value
1105
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1108
def check_legal_path(path):
1109
"""Check whether the supplied path is legal.
1110
This is only required on Windows, so we don't test on other platforms
1113
if sys.platform != "win32":
1115
if _validWin32PathRE.match(path) is None:
1116
raise errors.IllegalPath(path)
1119
def walkdirs(top, prefix=""):
1120
"""Yield data about all the directories in a tree.
1122
This yields all the data about the contents of a directory at a time.
1123
After each directory has been yielded, if the caller has mutated the list
1124
to exclude some directories, they are then not descended into.
1126
The data yielded is of the form:
1127
((directory-relpath, directory-path-from-top),
1128
[(relpath, basename, kind, lstat, path-from-top), ...]),
1129
- directory-relpath is the relative path of the directory being returned
1130
with respect to top. prefix is prepended to this.
1131
- directory-path-from-root is the path including top for this directory.
1132
It is suitable for use with os functions.
1133
- relpath is the relative path within the subtree being walked.
1134
- basename is the basename of the path
1135
- kind is the kind of the file now. If unknown then the file is not
1136
present within the tree - but it may be recorded as versioned. See
1138
- lstat is the stat data *if* the file was statted.
1139
- planned, not implemented:
1140
path_from_tree_root is the path from the root of the tree.
1142
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1143
allows one to walk a subtree but get paths that are relative to a tree
1145
:return: an iterator over the dirs.
1147
#TODO there is a bit of a smell where the results of the directory-
1148
# summary in this, and the path from the root, may not agree
1149
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1150
# potentially confusing output. We should make this more robust - but
1151
# not at a speed cost. RBC 20060731
1153
_directory = _directory_kind
1154
_listdir = os.listdir
1155
_kind_from_mode = _formats.get
1156
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1158
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1159
relroot, _, _, _, top = pending.pop()
1161
relprefix = relroot + u'/'
1164
top_slash = top + u'/'
1167
append = dirblock.append
1168
for name in sorted(_listdir(top)):
1169
abspath = top_slash + name
1170
statvalue = _lstat(abspath)
1171
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1172
append((relprefix + name, name, kind, statvalue, abspath))
1173
yield (relroot, top), dirblock
1175
# push the user specified dirs from dirblock
1176
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1179
def _walkdirs_utf8(top, prefix=""):
1180
"""Yield data about all the directories in a tree.
1182
This yields the same information as walkdirs() only each entry is yielded
1183
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1184
are returned as exact byte-strings.
1186
:return: yields a tuple of (dir_info, [file_info])
1187
dir_info is (utf8_relpath, path-from-top)
1188
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1189
if top is an absolute path, path-from-top is also an absolute path.
1190
path-from-top might be unicode or utf8, but it is the correct path to
1191
pass to os functions to affect the file in question. (such as os.lstat)
1193
fs_encoding = _fs_enc.upper()
1194
if (sys.platform == 'win32' or
1195
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1196
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1198
return _walkdirs_fs_utf8(top, prefix=prefix)
1201
def _walkdirs_fs_utf8(top, prefix=""):
1202
"""See _walkdirs_utf8.
1204
This sub-function is called when we know the filesystem is already in utf8
1205
encoding. So we don't need to transcode filenames.
1208
_directory = _directory_kind
1209
_listdir = os.listdir
1210
_kind_from_mode = _formats.get
1212
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1213
# But we don't actually uses 1-3 in pending, so set them to None
1214
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1216
relroot, _, _, _, top = pending.pop()
1218
relprefix = relroot + '/'
1221
top_slash = top + '/'
1224
append = dirblock.append
1225
for name in sorted(_listdir(top)):
1226
abspath = top_slash + name
1227
statvalue = _lstat(abspath)
1228
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1229
append((relprefix + name, name, kind, statvalue, abspath))
1230
yield (relroot, top), dirblock
1232
# push the user specified dirs from dirblock
1233
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1236
def _walkdirs_unicode_to_utf8(top, prefix=""):
1237
"""See _walkdirs_utf8
1239
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1241
This is currently the fallback code path when the filesystem encoding is
1242
not UTF-8. It may be better to implement an alternative so that we can
1243
safely handle paths that are not properly decodable in the current
1246
_utf8_encode = codecs.getencoder('utf8')
1248
_directory = _directory_kind
1249
_listdir = os.listdir
1250
_kind_from_mode = _formats.get
1252
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1254
relroot, _, _, _, top = pending.pop()
1256
relprefix = relroot + '/'
1259
top_slash = top + u'/'
1262
append = dirblock.append
1263
for name in sorted(_listdir(top)):
1264
name_utf8 = _utf8_encode(name)[0]
1265
abspath = top_slash + name
1266
statvalue = _lstat(abspath)
1267
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1268
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1269
yield (relroot, top), dirblock
1271
# push the user specified dirs from dirblock
1272
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1275
def copy_tree(from_path, to_path, handlers={}):
1276
"""Copy all of the entries in from_path into to_path.
1278
:param from_path: The base directory to copy.
1279
:param to_path: The target directory. If it does not exist, it will
1281
:param handlers: A dictionary of functions, which takes a source and
1282
destinations for files, directories, etc.
1283
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1284
'file', 'directory', and 'symlink' should always exist.
1285
If they are missing, they will be replaced with 'os.mkdir()',
1286
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1288
# Now, just copy the existing cached tree to the new location
1289
# We use a cheap trick here.
1290
# Absolute paths are prefixed with the first parameter
1291
# relative paths are prefixed with the second.
1292
# So we can get both the source and target returned
1293
# without any extra work.
1295
def copy_dir(source, dest):
1298
def copy_link(source, dest):
1299
"""Copy the contents of a symlink"""
1300
link_to = os.readlink(source)
1301
os.symlink(link_to, dest)
1303
real_handlers = {'file':shutil.copy2,
1304
'symlink':copy_link,
1305
'directory':copy_dir,
1307
real_handlers.update(handlers)
1309
if not os.path.exists(to_path):
1310
real_handlers['directory'](from_path, to_path)
1312
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1313
for relpath, name, kind, st, abspath in entries:
1314
real_handlers[kind](abspath, relpath)
1317
def path_prefix_key(path):
1318
"""Generate a prefix-order path key for path.
1320
This can be used to sort paths in the same way that walkdirs does.
1322
return (dirname(path) , path)
1325
def compare_paths_prefix_order(path_a, path_b):
1326
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1327
key_a = path_prefix_key(path_a)
1328
key_b = path_prefix_key(path_b)
1329
return cmp(key_a, key_b)
1332
_cached_user_encoding = None
1335
def get_user_encoding(use_cache=True):
1336
"""Find out what the preferred user encoding is.
1338
This is generally the encoding that is used for command line parameters
1339
and file contents. This may be different from the terminal encoding
1340
or the filesystem encoding.
1342
:param use_cache: Enable cache for detected encoding.
1343
(This parameter is turned on by default,
1344
and required only for selftesting)
1346
:return: A string defining the preferred user encoding
1348
global _cached_user_encoding
1349
if _cached_user_encoding is not None and use_cache:
1350
return _cached_user_encoding
1352
if sys.platform == 'darwin':
1353
# work around egregious python 2.4 bug
1354
sys.platform = 'posix'
1358
sys.platform = 'darwin'
1363
user_encoding = locale.getpreferredencoding()
1364
except locale.Error, e:
1365
sys.stderr.write('bzr: warning: %s\n'
1366
' Could not determine what text encoding to use.\n'
1367
' This error usually means your Python interpreter\n'
1368
' doesn\'t support the locale set by $LANG (%s)\n'
1369
" Continuing with ascii encoding.\n"
1370
% (e, os.environ.get('LANG')))
1371
user_encoding = 'ascii'
1373
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1374
# treat that as ASCII, and not support printing unicode characters to the
1376
if user_encoding in (None, 'cp0'):
1377
user_encoding = 'ascii'
1381
codecs.lookup(user_encoding)
1383
sys.stderr.write('bzr: warning:'
1384
' unknown encoding %s.'
1385
' Continuing with ascii encoding.\n'
1388
user_encoding = 'ascii'
1391
_cached_user_encoding = user_encoding
1393
return user_encoding
1396
def recv_all(socket, bytes):
1397
"""Receive an exact number of bytes.
1399
Regular Socket.recv() may return less than the requested number of bytes,
1400
dependning on what's in the OS buffer. MSG_WAITALL is not available
1401
on all platforms, but this should work everywhere. This will return
1402
less than the requested amount if the remote end closes.
1404
This isn't optimized and is intended mostly for use in testing.
1407
while len(b) < bytes:
1408
new = socket.recv(bytes - len(b))
1414
def dereference_path(path):
1415
"""Determine the real path to a file.
1417
All parent elements are dereferenced. But the file itself is not
1419
:param path: The original path. May be absolute or relative.
1420
:return: the real path *to* the file
1422
parent, base = os.path.split(path)
1423
# The pathjoin for '.' is a workaround for Python bug #1213894.
1424
# (initial path components aren't dereferenced)
1425
return pathjoin(realpath(pathjoin('.', parent)), base)
1428
def supports_mapi():
1429
"""Return True if we can use MAPI to launch a mail client."""
1430
return sys.platform == "win32"
added
removed
bzrlib/osutils.py
